Number Of Nearest Neighbors Research Articles

Explicit Incorporation of Spatial Autocorrelation in 3D Deep Learning for Geospatial Object Detection

Three-dimensional (3D) geospatial object detection has become essential for 3D geospatial studies driven by explosive growth in 3D data. It is extremely labor- and cost-intensive, though, as it often requires manual detection. Deep learning has been recently used to automate object detection within 3D context. Yet, addressing spatial dependency in 3D data and how it might inform deep learning for 3D geospatial object detection remains a significant challenge. Traditional models focus on the use of spatial properties, often overlooking color and contextual information. Exploiting these nonspatial attributes for 3D geospatial object detection thus becomes crucial. Our study pioneers explicit incorporation of spatial autocorrelation of color information into 3D deep learning for object detection. We introduce an innovative framework to estimate spatial autocorrelation, addressing challenges in unstructured 3D data sets. Our experiments suggest the effectiveness of incorporating spatial autocorrelation features in enhancing the accuracy of 3D deep learning models for geospatial object detection. We further investigate the uncertainty of such contextual information brought by diverse configurations, exemplified by the number of nearest neighbors. This study advances 3D geospatial object detection via using spatial autocorrelation to inform deep learning algorithms, strengthening the connection between GIScience and artificial intelligence and, thus, holding implications for diverse GeoAI applications.

G2-SCANN: Gaussian-kernel graph-based SLD clustering algorithm with natural neighbourhood

For most clustering methods, not only the number of clusters must be set in advance, but also various hyperparameters such as initial centroids, number of nearest neighbours, the minimum number of points, neighbourhood radius, and cutoff distance all require pre-specification. As one of the most promising unsupervised learning methods in machine intelligence, existing clustering methods cannot simultaneously handle datasets with arbitrary shapes, different densities, distinct sizes, and overlapping. Background outliers and high dimensionality make clustering problems more challenging. In this paper, we propose a novel universal clustering methodology, called G2-SCANN, which yields the best clustering performance for all 30 synthetic and real datasets without any hyperparameter tuning if the exact number of clusters is known. Firstly, the shortest path length (SPL) in complex network or graph-based geodesic distance is used to give a locally backbone-structured description of graph vertex similarity. Accordingly, SPL-weighted local degree (SLD) is defined as vertex attributes of a SPL-weighted graph expressed by G2-SPL adjacency matrix with ε-natural neighbourhood. Secondly, the process of calculating SLD for every data point in a bottom-up way directly leads to division from a complete graph constituted by all data points to a group of SLD trees. This brings the interpretability and the elimination of lone trees. Thirdly, contrastive learning of largest SLD values for finding root vertices of each divisive tree is conducted and top-down category message is then transmitted from the root vertices to all the leaf ones of a SLD tree. It eventually produces tree-like clusters. Totally, the proposed G2-SCANN method leverages both local neighbouring similarity of data points and global information about data distribution and makes it perform better than other methods.

On propensity score matching with a diverging number of matches

Abstract This paper re-examines the work of Abadie & Imbens (2016) on propensity score matching for average treatment effect estimation. We explore the asymptotic behaviour of these estimators when the number of nearest neighbours, M, grows with the sample size. It is shown, while not surprising, but technically nontrivial, that the modified estimators can improve upon the original fixed M-estimators in terms of efficiency. Additionally, we demonstrate the potential to attain the semiparametric efficiency lower bound when the propensity score admits some special structures, echoing the insight of Hahn (1998).

Structural modeling of high-entropy oxides battery anodes using x-ray absorption spectroscopy

High-entropy oxides (HEOs) are single phase solid solutions where five or more metals share the same sublattice, giving rise to unexpected features in various fields of applications. Recently, HEOs have emerged as an alternative conversion electrode anode material for next-generation Li-ion batteries, where the combination of several different elements in a single solid solution can synergistically act to overcome some of its main drawbacks, improving performance. Due to their chemical complexity, x-ray absorption spectroscopy (XAS) emerges as an appropriate technique to study the electronic (x-ray absorption near edge structure, XANES) and local structure (extended x-ray absorption fine structure, EXAFS) of these compounds as a function of cycling. This work aims to highlight the capabilities of XAS as an element-specific probe to understand a material’s structure at the atomistic level through EXAFS modeling of (MgFeCoNiCuZn)O high-entropy system and how to extract valuable information about the bond distance, number of near neighbors, and local disorder, which are crucial to a full understanding of the electrochemical reaction mechanisms of such battery electrodes.

Semi-automatic recognition of rock mass discontinuity based on 3D point clouds

Information on rock mass discontinuities is crucial for rock mass stability analysis. Due to the low efficiency, incompleteness, and potential risk of the traditional compass methods in measuring discontinuities, three-dimensional light detection, ranging, and other remote sensing methods have become essential. In this study, voxel filtering was used to subsample a point cloud so that its feature points were retained while reducing the computational load. An improved regional growing (RG) algorithm was then used to extract rock mass discontinuities. A software Geocloud v1.0 was developed based on the proposed method to semi-automatically recognize discontinuities. Additionally, two groups of sensitivity experiments were performed to analyze the influence of different numbers of nearest neighbors and maximum RG angles on the extraction of discontinuities. Results showed that most of the discontinuities could be accurately recognized with different thresholds. Furthermore, the accuracy of the proposed method was verified by real geometries, on a real highway slope, and in a natural quarry. Finally, the efficiency of the proposed method was proven using comparative computational experiments.

Monitoring Dynamically Changing Migratory Flocks Using an Algebraic Graph Theory-Based Clustering Algorithm

Migration flocks have different forms, including single individuals, formations, and irregular clusters. The shape of a flock can change swiftly over time. The real-time clustering of multiple groups with different characteristics is crucial for the monitoring of dynamically changing migratory flocks. Traditional clustering algorithms need to set various prior parameters, including the number of groups, the number of nearest neighbors, or the minimum number of individuals. However, flocks may display complex group behaviors (splitting, combination, etc.), which complicate the choice and adjustment of the parameters. This paper uses a real-time clustering-based method that utilizes concepts from the algebraic graph theory. The connected graph is used to describe the spatial relationship between the targets. The similarity matrix is calculated, and the problem of group clustering is equivalent to the extraction of the partitioned matrices within. This method needs only one prior parameter (the similarity distance) and is adaptive to the group’s splitting and combination. Two modifications are proposed to reduce the computation burden. First, the similarity distance can be broadened to reduce the exponent of the similarity matrix. Second, the omni-directional measurements are divided into multiple sectors to reduce the dimension of the similarity matrix. Finally, the effectiveness of the proposed method is verified using the experimental results using real radar data.

Connectivity of stormwater ponds impacts Odonata abundance and species richness

ContextThe successful dispersal of an animal depends, partly, on landscape connectivity. Urbanization poses risks to dispersal activities by increasing hostile land cover types.ObjectivesWe investigated how connectivity of urban ponds impacted Odonata communities (dragonflies and damselflies), an order of semi-aquatic insects that actively disperse.MethodsWe sampled 41 constructed stormwater ponds and 8 natural ponds in a metropolitan area. The effect of connectivity and the quantity of available adjacent habitats was tested at different scales for dragonflies (900 m) and damselflies (300 m), determined by a literature analysis, to account for differences in suborder dispersal capabilities.ResultsLower levels of connectivity and fewer nearest neighbours negatively impacted abundance, species richness, and composition of dragonflies (p values < 0.01, R2 = 0.18–0.70). Adult dragonfly abundance had a stronger positive relationship with connectivity than species richness. In particular, the abundance of adult dragonfly Leucorrhinia frigida, found almost exclusively at natural ponds, had a positive relationship with connectivity. Connectivity and the number of nearest neighbours had no significant impact on damselflies apart from a slight negative relationship between connectivity and species richness (p value = 0.02, R2 = 0.11). Natural ponds had significantly higher levels of connectivity when compared to stormwater ponds.ConclusionsOur results suggest that dragonflies are positively affected by increased connectivity in an urban landscape, with no benefit of connectivity to damselflies at the scale measured. We recommend intentional planning of urban stormwater pond networks, where individual ponds can act as stepping stones, incorporated with strategic inclusion of beneficial land cover types.

Pairing-specific microstructure in depletion gels of bidisperse colloids.

We report the ensemble-averaged and pairing-specific network microstructure formed by short-range depletion attractions in hard sphere-like colloidal systems. Gelation is induced by adding polystyrene molecules at a fixed concentration to colloids with different colloid bidispersity ratios (α = 1, 0.72, and 0.60) across a range of volume fractions (0.10 ≤ ϕ ≤ 0.40). 3D confocal microscopy imaging combined with a scale-invariant feature transform algorithm show that monodisperse colloids pack more efficiently, whereas increasing the size disparity leads to looser, more disordered, and sub-isostatic packings. Categorizing the structures formed by small and large particles reveal that certain cluster configurations may be favored due to the complex interplay between the differences in particle surface areas and attractive potentials. These pairwise bonds assemble to affect the density of tetrahedral and poly-tetrahedral clusters in bidisperse systems. With the exception of non-percolating samples at ϕ = 0.10, increasing the gel volume fraction leads to an increase in the number of nearest neighbors. However, the internal density within each cluster decreases, possibly due to kinetic arrest from the deeper potential wells of tetrahedral clusters at low volume fractions in which vertices are primarily made out of larger particles.

Whale Optimization-based Synthetic Minority Oversampling Technique for Binary Imbalanced Datasets

The problem of class imbalance has become a predominant area of research recently. Synthetic Minority Oversampling Technique (SMOTE) stands as a popular and widely adopted oversampling technique that effectively addresses the challenge of class imbalance. However, its performance relies on one of the critical parameters i.e., the number of nearest neighbors, k_neighbors, which is often arbitrarily chosen by the users due to which it may not yield optimal results. Furthermore, the varying imbalance ratios across datasets add complexity to the task of parameter selection in SMOTE. In an effort to address this issue, this paper proposes a hybrid rebalancing technique called Whale optimization algorithm based SMOTE (WOA-SMOTE) that combines a metaheuristic technique, WOA, with SMOTE. The algorithm utilizes the advantages of WOA in finding the optimal value of k_neighbors of SMOTE which is crucial in generating synthetic samples that represent the distribution of samples more appropriately, thereby improving the performance of classifiers on imbalanced datasets. The study evaluates the performance of WOA-SMOTE alongside 6 benchmark sampling techniques on 10 real-world imbalanced datasets from Keel repository. These belong to different domains with imbalance ratio (IR) ranging from 1.25 to 15.46. Four different classifiers are used and the evaluation is based on three performance measures: AUC, g-mean and F1 scores. The experimental results showcase WOA-SMOTE's superior performance over SMOTE in majority of the datasets. Notably, WOA-SMOTE outperforms existing techniques in terms of F1 scores when using SVM and XGBoost classifiers in 8 out of 10 datasets. Moreover, its performance remains impressive in 6 of these datasets with random forest and logistic regression classifiers.

CHIP-KNNv2: AConfigurable andHigh-PerformanceK-NearestNeighbors Accelerator on HBM-based FPGAs

The k-nearest neighbors (KNN) algorithm is an essential algorithm in many applications, such as similarity search, image classification, and database query. With the rapid growth in the dataset size and the feature dimension of each data point, processing KNN becomes more compute and memory hungry. Most prior studies focus on accelerating the computation of KNN using the abundant parallel resource on FPGAs. However, they often overlook the memory access optimizations on FPGA platforms and only achieve a marginal speedup over a multi-thread CPU implementation for large datasets.In this article, we design and implement CHIP-KNN: an HLS-based, configurable, and high-performance KNN accelerator. CHIP-KNN optimizes the off-chip memory access on modern HBM-based FPGAs such as the AMD/Xilinx Alveo U280 FPGA board. CHIP-KNN is configurable for all essential parameters used in the algorithm, including the size of the search dataset, the feature dimension and data type representation of each data point, the distance metric, and the number of nearest neighbors - K. In terms of design architecture, we explore and discuss the tradeoffs between two design versions: CHIP-KNNv1 (Ping-Pong buffer based) and CHIP-KNNv2 (streaming-based). Moreover, we investigate the routing congestion issue in our accelerator design, implement hierarchical structures to shorten critical paths, and integrate an open-source floorplanning optimization tool called TAPA/AutoBridge to eliminate the place-and-route issues. To explore the design space and balance the computation and memory access performance, we also build an analytical performance model. Given a user configuration of the KNN parameters, our tool can automatically generate TAPA HLS C code for the optimal accelerator design and the corresponding host code, on the HBM-based FPGA platform.Our experimental results on the Alveo U280 show that, compared to a 48-thread CPU implementation, CHIP-KNNv2 achieves a geomean performance speedup of 15×, with a maximum speedup of 45×. Additionally, we show that CHIP-KNNv2 achieves up to 2.1× performance speedup over CHIP-KNNv1 while increasing configurability. Compared with the state-of-the-art Facebook AI Similarity Search (FAISS) [23] GPU implementation running on a Nvidia Tesla V100 GPU, CHIP-KNNv2 achieves an average latency reduction of 30.6× while requiring 34.3% of GPU power consumption.

Adaptive evidential K-NN classification: Integrating neighborhood search and feature weighting

The number of nearest neighbors K and the utilized distance measure considerably impact the performance of the K-nearest neighbor (K-NN) algorithm. The information provided by neighbors may be perplexing due to the set presumption of K for every test sample without any prior knowledge, which might result in incorrect classification results. The hypothesis of locally constant class conditional probabilities relied on by K-NN, is violated by high dimensions, which also bring out the dimensionality curse. Notably, in the case of imperfect data, training samples may be noise-corrupted or located in substantially overlapping regions, thus impairing the effectiveness of the K-NN rule. In contrast to existing research, which frequently focuses on just one of the problems above, this paper presents an adaptive evidential K-NN classification (AEK-NN) algorithm that synchronizes neighborhood search with feature weighting. In order to jointly learn the adaptive neighborhood for each sample and various feature weights, AEK-NN maps the Euclidean space into a rebuilt space consisting of similarity coefficients through a two-stage training process. Within the evidence theory framework, AEK-NN produces classification results for data with imperfect labels. The advantages of applying evidence theory, incorporating adaptive neighborhood search, and feature weighting are shown through ablation studies. We also use simulated and real-world datasets to demonstrate the state-of-the-art performance of AEK-NN.

Analysis of media content recommendation in the new media era considering scene clustering algorithm

Abstract With the advent of the new media era, we face the problem of information overload every day, based on which this paper proposes a scenario-conscious clustering algorithm for recommending media content. The improved K-means algorithm is used to cluster the media content, an initial cluster center is randomly selected, and the remaining initial cluster centers are obtained by executing the improved algorithm to reduce the number of iterations and avoid neighboring situations. The clustering algorithm is then compared with content-based recommendation techniques, KNN recommendation algorithm, and collaborative filtering recommendation algorithm in terms of accuracy, recall, MAE value, and execution time. The clustering algorithm is significantly better than the other algorithms regarding accuracy and recall. The accuracy of the clustering algorithm is 0.4 when the recommendation sequence is 30, which is 0.4 higher than the collaborative filtering technique, 0.1 higher than the KNN, and 0.12 higher than the content recommendation method. In terms of MAE value, the clustering algorithm outperforms the other algorithms when the number of nearest neighbors is selected to be above 20. In terms of execution time, the longer the amount of data, the more obvious the advantage of the clustering algorithm. Therefore, the applicability and reliability of the model proposed in this paper for media content recommendation are verified in terms of accuracy, recall and execution time, which meet the design requirements.

A novel double-hybrid learning method for modal frequency-based damage assessment of bridge structures under different environmental variation patterns

Monitoring of modal frequencies under an unsupervised learning framework is a practical strategy for damage assessment of civil structures, especially bridges. However, the key challenge is related to high sensitivity of modal frequencies to environmental and/or operational changes that may lead to economic and safety losses. The other challenge pertains to different environmental and/or operational variation patterns in modal frequencies due to differences in structural types, materials, and applications, measurement periods in terms of short and/or long monitoring programs, geographical locations of structures, weather conditions, and influences of single or multiple environmental and/or operational factors, which may cause barriers to employing state-of-the-art unsupervised learning approaches. To cope with these issues, this paper proposes a novel double-hybrid learning technique in an unsupervised manner. It contains two stages of data partitioning and anomaly detection, both of which comprise two hybrid algorithms. For the first stage, an improved hybrid clustering method based on a coupling of shared nearest neighbor searching and density peaks clustering is proposed to prepare local information for anomaly detection with the focus on mitigating environmental and/or operational effects. For the second stage, this paper proposes an innovative non-parametric hybrid anomaly detector based on local outlier factor. In both stages, the number of nearest neighbors is the key hyperparameter that is automatically determined by leveraging a self-adaptive neighbor searching algorithm. Modal frequencies of two full-scale bridges are utilized to validate the proposed technique with several comparisons. Results indicate that this technique is able to successfully eliminate different environmental and/or operational variations and correctly detect damage.

BPF: a novel cluster boundary points detection method for static and streaming data

Data points situated near a cluster boundary are called boundary points and they can represent useful information about the process generating this data. The existing methods of boundary points detection cannot differentiate boundary points from outliers as they are affected by the presence of outliers as well as by the size and density of clusters in the dataset. Also, they require tuning of one or more parameters and prior knowledge of the number of outliers in the dataset for tuning. In this research, a boundary points detection method called BPF is proposed which can effectively differentiate boundary points from outliers and core points. BPF combines the well-known outlier detection method Local Outlier Factor (LOF) with Gravity value to calculate the BPF score. Our proposed algorithm StaticBPF can detect the top-m boundary points in the given dataset. Importantly, StaticBPF requires tuning of only one parameter i.e. the number of nearest neighbors (k) and can employ the same k used by LOF for outlier detection. This paper also extends BPF for streaming data and proposes StreamBPF. StreamBPF employs a grid structure for improving k-nearest neighbor computation and an incremental method of calculating BPF scores of a subset of data points in a sliding window over data streams. In evaluation, the accuracy of StaticBPF and the runtime efficiency of StreamBPF are evaluated on synthetic and real data where they generally performed better than their competitors.

Akaike information criterion in choosing the optimal k-nearest neighbours of the spatial weight matrix

ABSTRACT We use the Akaike information criterion (AIC) to assess the quality of non-nested spatial econometric models with a different number of nearest neighbours (knn) included in the spatial weight matrix W. This is of particular importance in two cases: when estimating the model on geolocated point data without explicit guidance on the optimal value of knn; and when dealing with different spatial patterns and a unique W generalizes diverse spatial structures. By minimizing AIC for a set of competing models, one can find the optimal knn that guarantees the best fit. We show through simulation and empirical analysis that AIC is a non-linear function of knn and reaches its minimum for a given knn. We provide practical evidence that misspecified W may result in a 20% bias of model parameters.

WordNet Semantic Relations Based Enhancement of KNN Model for Implicit Aspect Identification in Sentiment Analysis

Opinion mining or sentiment analysis (SA) is a key component of real-world applications for e-commerce organizations, manufacturers, and customers. SA deals with the computational evaluation of people’s views, thoughts, and feelings in the text, whether they are visible or concealed. The Aspect based SA level is becoming one of the most active phases in this area. In this paper, we propose an approach to enrich K-Nearest Neighbors (KNN) to deal with Implicit Aspect Identification task (IAI). Through the use of WordNet semantic relations, we propose an enhancement for KNN distance computation to support the IAI task. For a conclusive empirical evaluation, experiments are conducted on two datasets of electronic products and restaurant reviews and the effects of our approach are examined and analyzed according to three criteria: KNN distance used to compute the similarity, the number of nearest neighbors (K) and the KNN behavior towards Overfitting and Underfitting. The experimental results show that our approach helps KNN improve its performance and better deal with Overfitting and Underfitting for Implicit Aspect Identification.

A new hybrid approach based on genetic algorithm and support vector machine methods for hyperparameter optimization in synthetic minority over-sampling technique (SMOTE)

&lt;abstract&gt; &lt;p&gt;The crucial problem when applying classification algorithms is unequal classes. An imbalanced dataset problem means, particularly in a two-class dataset, that the group variable of one class is comparatively more dominant than the group variable of the other class. The issue stems from the fact that the majority class dominates the minority class. The synthetic minority over-sampling technique (SMOTE) has been developed to deal with the classification of imbalanced datasets. SMOTE algorithm increases the number of samples by interpolating between the clustered minority samples. The SMOTE algorithm has three critical parameters, "k", "perc.over", and "perc.under". "perc.over" and "perc.under" hyperparameters allow determining the minority and majority class ratios. The "k" parameter is the number of nearest neighbors used to create new minority class instances. Finding the best parameter value in the SMOTE algorithm is complicated. A hybridized version of genetic algorithm (GA) and support vector machine (SVM) approaches was suggested to address this issue for selecting SMOTE algorithm parameters. Three scenarios were created. Scenario 1 shows the evaluation of support vector machine SVM) results without using the SMOTE algorithm. Scenario 2 shows that the SVM was used after applying SMOTE algorithm without the GA algorithm. In the third scenario, the results were analyzed using the SVM algorithm after selecting the SMOTE algorithm's optimization method. This study used two imbalanced datasets, drug use and simulation data. After, the results were compared with model performance metrics. When the model performance metrics results are examined, the results of the third scenario reach the highest performance. As a result of this study, it has been shown that a genetic algorithm can optimize class ratios and k hyperparameters to improve the performance of the SMOTE algorithm.&lt;/p&gt; &lt;/abstract&gt;

ПОРІВНЯННЯ АЛГОРИТМІВ КЛАСИФІКАЦІЇ BIG DATA МЕТОДАМИ ІМІТАЦІЙНОГО МОДЕЛЮВАННЯ

With the development of information transmission and storage technologies, the volumes of data that require processing and analysis are growing rapidly. Therefore, the task of developing algorithms for solving various artificial intelligence problems for Big Data volumes is urgent. In our works, this informal term "Big Data" refers to situations when known processing algorithms do not allow solving a problem in a practically acceptable time. With regard to classification tasks, such conditions are possible when the first place is not even high reliability (that is, the minimum number of errors), but productivity (classification speed). The well-known method of nearest neighbors is one of the most productive. However, the indicator of the order of growth (the number of typical operations) for it is K x M x N, where K is the number of nearest neighbors, M is the number of classes, N is the typical number of class elements. Along with this, we propose to consider algorithms based on the principles of M-means, where classes are replaced by only a small number of their characteristics. Among such algorithms, the article considers: the algorithm of class centers and the algorithm of adaptive rules. The order of growth for these algorithms is only M according to the number of classes. The comparative analysis of these algorithms is performed by the method of simulation modeling. Simulation models are implemented by the Adaptive Metrics program, developed at the Department of Software Engineering at DUITZ. In this program, the classification problem is solved using the example of the dichotomy problem for classes A and B. The program has the possibility of very flexible setting of models. Problems can be solved in 1-dimensional, 2-dimensional,..., 6-dimensional spaces. The distribution of factor values for classes A and B can have quite different statistical characteristics - from uniform and triangular distribution functions to functions approaching a normal distribution. The graphical interface of the program allows you to dynamically observe the solution of the classification problem in one-dimensional, two-dimensional and 6-dimensional projections. As a result of multiple runs of the program, it was established that the algorithms of the nearest neighbors slightly outperform the algorithms of class centers and adaptive rules according to the criterion of reliability, and also comply with the principle of compactness (concentration of the largest number of erroneous solutions in the hypercube of errors). Algorithms based on M-means principles significantly outperform this algorithm in terms of performance. Also, the algorithm of adaptive rules best corresponds to the principle of equality of classes and is the most productive of the considered ones.

An optimized case-based software project effort estimation using genetic algorithm

Software development companies have long suffered from inaccurate estimation of their software projects. This in turn led to huge losses, especially in the financial resources available for the project as well as the time required to complete it. As a result of this, the research community has developed different methods for estimating effort in software projects in the hope of achieving high levels of accuracy and efficiency in the use of available resources. Among those methods that have proven to be accurate in estimating the effort of software projects is the use of machine learning (ML) techniques, especially the case-based reasoning technique (CBR). This technique is based on adapting previously successful solutions for similar software projects. However, the CBR technique suffers from a problem which is its multiple parameters that are difficult to be tuned. This justifies the importance of the adaptation and adjustment process as an essential part of CBR to produce accurate and efficient results with least absolute estimation error. In this paper, one of the most efficient multi-objective evolutionary techniques, the Genetic Algorithm (GA), are used to help find the best set of classical CBR parameters (feature selection, feature weighting, similarity measures, and k number of nearest neighbors) to produce the most accurate effort estimates for software projects. The proposed CBR-GA model showed the effectiveness of using the GA algorithm to search for the best combination of CBR parameters and thus improve its accuracy. This in turn is beneficial for project managers in the early financial planning phase for effort estimation and thus project cost control. To validate the proposed CBR-GA model, we used a set of public benchmark datasets available on PROMISE data repository, in addition we used a set of reliable evaluation metrics. The obtained results are promising in terms of accuracy and significance tests. This implies the importance of search-based techniques for tuning effort estimation methods.

Estimation Based on Nearest Neighbor Matching: From Density Ratio to Average Treatment Effect

Nearest neighbor (NN) matching is widely used in observational studies for causal effects. Abadie and Imbens (2006) provided the first large‐sample analysis of NN matching. Their theory focuses on the case with the number of NNs, M fixed. We reveal something new out of their study and show that once allowing M to diverge with the sample size an intrinsic statistic in their analysis constitutes a consistent estimator of the density ratio with regard to covariates across the treated and control groups. Consequently, with a diverging M, the NN matching with Abadie and Imbens' (2011) bias correction yields a doubly robust estimator of the average treatment effect and is semiparametrically efficient if the density functions are sufficiently smooth and the outcome model is consistently estimated. It can thus be viewed as a precursor of the double machine learning estimators.